• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    Improvements of Vis-NIRS Model in The Prediction of Soil Organic Matter Content Using Wavelength Optimization

    2016-12-15 05:04:59LINZhidanWANGYubingWANGRujingWANGLiusanLUCuipingZHANGZhengyongSONGLiangtuLIUYang
    發(fā)光學(xué)報 2016年11期
    關(guān)鍵詞:中科院合肥波長

    LIN Zhi-dan, WANG Yu-bing, WANG Ru-jing, WANG Liu-san,LU Cui-ping, ZHANG Zheng-yong, SONG Liang-tu, LIU Yang

    (1. Institute of Intelligent Machines, Chinese Academy of Sciences, Hefei 230031, China;2. Department of Automation, University of Science and Technology of China, Hefei 230026, China;3. Electronic Engineering Institute, Hefei 230037, China)

    ?

    Improvements of Vis-NIRS Model in The Prediction of Soil Organic Matter Content Using Wavelength Optimization

    LIN Zhi-dan1,2,3, WANG Yu-bing1*, WANG Ru-jing1*, WANG Liu-san1,LU Cui-ping1, ZHANG Zheng-yong1, SONG Liang-tu1, LIU Yang1

    (1.InstituteofIntelligentMachines,ChineseAcademyofSciences,Hefei230031,China;2.DepartmentofAutomation,UniversityofScienceandTechnologyofChina,Hefei230026,China;3.ElectronicEngineeringInstitute,Hefei230037,China)

    Visible-near infrared spectroscopy (Vis/NIRS) is proved to be an effective tool in the prediction of soil properties. Wavelength optimization plays an important role in the construction of Vis-NIRS prediction model. In this article, a total of 130 topsoil samples collected from Guoyang County, Anhui Province, China, were used to establish a Vis-NIRS model for the prediction of organic matter content (OMC) in line concretion black soils. Through comparison, the combined spectral pretreatments of smooth and multiplicative scatter correlation (MSC) were applied to minimize the irrelevant and useless information of the spectra and increase the correlation between spectra and the measured values, and subsequently, SPXY methods were used to select the representative training set. Successive projection algorithm (SPA) and genetic algorithm (GA) were then conducted for wavelength optimization. Finally, the principal component regression (PCR) model was constructed, in which the optimal number of principal components was determined using leave-one-out cross validation technique. Results show that: both SPA and GA can significantly reduce the wavelength and favorably increase the accuracy, especially, GA can greatly improve the prediction accuracy of soil OMC, withRcc, RMSEP and RPD up to 0.931 6, 0.214 2, 2.319 5, respectively. Conclusively, using appropriate wavelength optimization methods, not only the computational load can be significantly reduced but also the prediction precision can be improved.

    visible-near infrared spectroscopy(Vis-NIRS); organic matter content (OMC); spectral pretreatments; sample selection; wavelength optimization

    1 Introduction

    The application of precision agriculture needs vast amounts of accurate, real-time and low-cost soil data over a large area, which sets high requirements for soil detection methods. The detection method should be more efficient in both detection time and cost. Visible and near infrared reflectance spectroscopy (Vis-NIRS) is a non-destructive, rapid and repeatable method that can serve as an effective substitute technology for traditional laboratory chemical detection methods[1-4]. When a sample is illuminated by light, certain bonds in the molecules vibrate with the varying electric field, which will absorb optical energy and cause less light to be reflected off the sample. Three of the most pronounced vibrations observable in near-infrared band are the groups involving C—H, N—H and O—H bonds, which thus makes it possible to measure soil properties such as moisture, organic C and N using Vis-NIRS technique[5-6]. Each soil property has distinct spectral fingerprints in Vis-NIR region. Owing to the significant developments in equipment, optical components, computer and chemometrics, Vis-NIRS now is extensively applied in many fields for qualitative and quantitative analyses[7].

    Soil organic matter content (OMC) plays a major role in soil’s many chemical and physical processes and significantly affects the pattern of a soil reflectance spectrum. As to the prediction of OMC using Vis-NIRS technique, many researchers have conducted a great number of studies and achieved many favorable results. Based on the collected Vis-NIR data, Vasquesetal.[8-9], Mouazenetal.[10], Stevensetal.[11], and Rossel and Behrens[12]used different multivariate regression methods and systematically compared their capabilities in soil OMC contents. Principle component regression (PCR) is a common model in multiple linear regression. Changetal. evaluated the ability of near infrared spectroscopy to predict soil OMC with PCR, with the determination coefficient up to 0.87 and the residual prediction deviation (RPD) of 2.79[13]. Vasquesetal. identified the best combination to predict soil OMC with five multivariate techniques including PCR[8].Wangetal. analyzed the potential of Vis-NIRS to predict soil OMC using two spectrometers, and the results showed that both two spectrometers can achieve favorable results in the prediction of soil OMC[14]. All of those proved the feasibility of PCR-based NIRS model in the prediction of OMC. However, the noise and some irrelevant or collinear information concluded in Vis-NIRS can affect the accuracy of the PCR method, which should be eliminated with some measures before the establishment of model.

    In this paper, PCR was used to relate the Vis/NIR spectra with soil OMC, while spectral pretreatments, sample selection and wavelength optimization were conducted for improving the prediction accuracy of the constructed model. By comparing the prediction results using different methods, the application of spectral pretreatment, sample selection and wavelength optimization on the improvement of soil OMC prediction capacities were evaluated.

    2 Materials and Methods

    2.1 Collection of Soil Samples

    The experimental field in Tongfeng Seed Industry was selected in the present study, which is located in the Guoyang County, Bozhou City, Anhui Province, China (33°27′-33°47′N; 115°53′-116°33′E). The mean annual temperature is about 14.6 ℃, and the mean annual precipitation is about 830 mm. The overall flat fields in Guoyang County can be classified as the lime concretion black soil. As one type of ancient cultivated soils, lime concretion black soil presents a highly localized distribution in Huaibei Plain, China. According to China Soil Scientific Database (http://www.soil.csdb.cn/), it is composed of two layers from top to bottom, namely, black-soil layer and lime-concretion layer, respectively. Generally, the lime concretion black soils are abundant in K content but relatively poor in OM, N and P contents, which can also be reflected by the data in Tab.1[15-16].

    A total of 130 topsoil samples were collected using S-type sampling. Soil samples were collected from the surface layer at the bottom of a 20 cm deep trench, with the use of a special soil sampler. Each sample was about 2 000 g and was placed into a tightly sealed plastic bag to avoid external contamination. After the rejection of weeds and small pieces of rocks, the soils were then naturally air-dried and sieved less than 2 mm[17]. The soil samples after preprocessing were divided into two portions by four-way division method: A and B. The samples in Group A were placed in properly closed bags and taken to the chemical laboratory for the analysis of OMC. The soil OMC was determined by the potassium dichromate volumetric method coupling with a watering heating technique[18]. The statistical data of measured results are listed in Tab.1. The samples in Group B were taken to spectral measurements, which are described in depth in the next section.

    Tab.1 Statistical data of all samples measured results

    2.2 Spectral Data Acquisition

    The experimental instrument employed in the present work, was Vis/NIR soil sensor by Veris Technology Incorporation. As a tractor-mounted sensor that can collect real-time soil information, the spectrophotometers and the optical system of Veris Vis/NIR soil sensor were built into a shank, mounted on a toolbar and then pulled by a tractor during field investigations. In present work, the spectrophotometers (Ocean Optics USB4000 and Hamamatsu C9914GB) and the optical system (a tungsten halogen bulb and fibers) in Veris Soil Sensor were dismounted from the tractor-mounted mobile platform and performed measurements in laboratory. The total spectra range from 342 to 2 222 nm, which were automatically stitched by the software at the absorption terminal. All the data processing procedures in present work were compiled with Matlab. A total of 130 samples was adopted for measurements, which were put in a petri dish and the surface was smoothed beforehand. During the measurements, the sample surface was pressed against the sapphire window of Veris soil sensor. To make a tradeoff between the minimization of the measured errors and time consumptions, each sample was detected three times, which was rotated by 120° for the next scan. After each measurement, the collected three spectra was averaged. The averaged absorbance soil spectra was given in Fig.1, in which they-axis absorbance spectra (A) were converted from reflectance spectra (R) byA=log10(1/R) and thex-axis wavelength was reciprocal to wave number. It can be found that the first two large absorption peaks are located at around 1 420 and 1 930 nm, respectively, both of which are coincident with the characteristic absorption peaks of H2O[18].

    Fig.1 Vis-NIR spectra of 130 samples measured by the Veris soil sensor

    2.3 Spectral Pretreatments

    The measured spectra were easily influenced by individual differences (the particle size of samples, the intensity of light, the condition of measurement,etc.), baseline variations and substantial noises. Therefore, the pretreatment should be applied to minimize the irrelevant and useless information of the spectra and increase the correlation between the spectra and the measured values. The frequently-adopted pretreatment methods include normalization, first and second derivatives, multiplicative scatter correlation (MSC), standard normal variate (SNV), detrending or any combination thereof[19-20]. In the present study, the pretreatment and the subsequent processing programs were compiled with MATLAB 2012b.

    SNV and MSC transformations can remove the baseline drift from spectra caused by the scattering and the variation of particle sizes. To remove high frequency noise, Savitzky-Golay polynomial smoothing filter can digitally smooth a given spectrum by approximating it within a specified data window using a polynomial with a specified order. Accordingly, the data can be best matched in the window on a least-square basis. In the present work, the filter with a polynomial of order 3 and the window with the width of 7 data points were used[14]. By comparing the results using 17 different pretreatment methods, the best pretreatment method employed in present work was the combination of S-G filter for smoothing and MSC.

    2.4 Sample Selection

    The selection of a representative training set plays a determinative role in the construction of prediction models, since the models established with the representative-characteristics samples can lead to the acceleration of regressions, the improvement of the prediction accuracy and the reduction of storage space and costs. Moreover, the application range of the established models can be expanded by adding a small amount of representative samples, being beneficial to the update and improvement of the models.

    Kennard-Stone (KS) method and the sample set partitioning based on jointx-ydistance (SPXY) method Kennard-Stone (K-S) algorithm aim at covering the multidimensional space in a uniform manner by maximizing the Euclidean distances between the instrumental response vectors (x) of the selected samples[21]. SPXY method extends the K-S algorithm by encompassing bothx- andy- differences in the calculation of inter-sample distances[22].

    2.5 Wavelength Optimization

    Wavelength optimization on the full spectrum with the aim of enhancing accuracy is still a challenging task, especially when the collected spectra display strong overlapping and imperceptible distinctive features. The spectra in Vis-NIR range are mainly composed of the overtones and combination bands of hydrogen groups, and the absorption peaks are of weak intensity, relatively low sensitivity, wide absorption band width, serious overlaps and multiple correlations in spectral information. If the full spectrum was involved in the model, it would not only increase the complexity of the model and calculation load, but also reduce the prediction accuracy of the model owing to the irrelevant variables and collinearity between variables.

    Successive projections algorithm (SPA) selects the wavelengths according to the contribution value sequence of the test samples and looks for the original spectral data with minimum redundant information. Accordingly, the overlapping information can be avoided and the redundant information can be eliminated in the selected wavelength data. This method can greatly reduce the amount of calculation model and improve the stability and accuracy of the model[23].

    Genetic algorithm (GA) is a kind of random search optimization algorithm by reference to the rule of biological evolution. Owing to the invisible parallelism, adaptive and global optimization ability, GA has become a common method for the optimization of wavelength in the construction of NIR prediction models. In combination with GA algorithm, the constructed NIR prediction models exhibit relatively high predictive abilities[24].

    2.6 Calibration and Validation

    Among the training set after selection with 130 samples, the calibration and validation sets were selected in the present work, with the aim of prediction an unknown sample scientifically and exactly. The selected 100 samples were divided as calibration data set, and the rest 30 samples were used as the validation set. In calibration stage, the spectra were compressed using principal components analysis (PCA), and the optimum number of principal components (PCs) was determined using leave-one-out cross validation (LOOCV) technique, in which each sample was omitted and predicted using the calibration model established by the remaining samples[25]. With the pre-processed spectra, the prediction residual error sum of squares (PRESS) in leave-one-out cross validation for different number of PCs and soil contents was calculated. A suitable number of PCs is an efficient way of taking full advantage of spectral information and noise-filtering, while some useless information, such as the measured errors, can be over-included for more number of PCs, also known as ‘over-fitting’. Subsequently, the models were constructed using PCR method, in which multiple linear regressions were performed using the obtained optimum numbers of PCs. In validation stage, the above-described calibration model developed from training set (100 samples) was used to predict the contents of soil samples in validation set (30 samples), and the predicted values were compared with measured values. The statistic parameters for evaluating the predictive capability of the models include correlation coefficient (Rcc), root means square error of prediction (RMSEP) and the ratio of standard deviation of the validation set to standard error of prediction (RPD)[26].

    3 Results and Discussion

    The spectra after the combined pretreatments of S-G filter for smoothing are displayed in Fig.2. One

    can observe that, compared with the original spectra, the pretreated spectra can smooth the spectra, reduce the effect of noise and then enhance the spectral characteristics.

    Fig.2 Vis-NIR spectra of 130 samples after the combined pretreatment of S-G filter for smoothing and MSC

    After the pretreatments, SPXY and KS method were used for the selection of training set and PCR-based prediction models for OMC were then developed. Tab.2 and Fig.3 show the results of OMC prediction using different training set selection methods. It can be observed that, using random selection (RS), the prediction results are considerably poor; the prediction results using KS sample selection are much better than the results using RS, asRccincreases from 0.329 0 to 0.719 8, RMSEP decreases from 0.453 7 to 0.437 9, and RPD increases from 1.078 0 to 1.282 6. The prediction results using SPXY results are best, withRcc, RMSEP and RPD up to 0.829 8, 0.273 8, 1.721 6, respectively. As stated above, SPXY method selects the training set by calculating the inter-sample distances based on overall considerations of the NIR spectral and target variables, so the selected training set is more representative than that selected using KS method and the established model is best in prediction accuracy. Consequently, SPXY method was picked out for further in-depth discussions in calibration and validation.

    Tab.2 OMC prediction results of PCR models with smooth+MSC and different sample selection methods

    Fig.3 Correlation between the predicted and the measured values of OMC with different sample selection methods. (a) RS. (b) KS. (c) SPXY.

    Tab.3 lists the results of PCR models with the use of different wavelength optimization methods (as the modeling detains illustrated in Section 2.6), and Fig.4 shows the comparison between the predicted results and the measured values. As shown in Tab.3, SPA and GA can both increase the accuracy of PCR model and greatly decrease the predicting errors. Both two methods can contribute to the optimization of wavelengths so as to remove the effects of noise and enhance the predictive capability. This can be reflected by the increased correlation coefficientRccand RPD as well as the decreased RMSEP.

    Fig.5 displays the characteristic wavelengths selected by SPA and GA, respectively. SPA and GA employ simple operations in a vector space to obtain the subsets of variables with small collinearity, which can effectively eliminate the redundant information of the wavelength variables and thus improve

    Tab.3 Prediction results of PCR models with different wavelength optimization methods

    Fig.4 Correlation between the predicted and the measured values of OMC with different wavelength selection methods. (a) NULL. (b) SPA. (c) GA.

    the prediction precision. In the present study, using SPA method, the optimal wavelength combination was selected based on the calculated RMSEP.While using GA method, the number of iterations, the population size, the genetic probability and mutation probability were set as 250, 30, 0.7 and 0.3, respectively. Specifically, SPA requires a less computational work load than GA, however, GA-based prediction model has a better prediction precision. Since GA is an adaptive global search algorithm and takes the serial correlation characteristics of the spectral data into account in the optimization of wavelength, more useful spectral information was included while some irrelevant or nonlinear variables were rejected. Thus, the GA-PCR model is superior to SPA-PCR model in the prediction of soil OMC.

    4 Conclusion

    In order to establish an accurate and robust prediction model for soil OMC, different pretreatment methods, sample selection methods and wavelength optimization methods were applied in the process of Vis-NIRS model establishment. Results show that the combination of S-G filter for smooth and MSC can effectively eliminate the effects of noise and baseline drift better. Both KS and SPXY can select the representative samples, but SPXY method overall considers bothx-andy-differences in the calculation of inter-sample distances, and can select the more representative samples and gain more accurate results. SPA and GA can decrease the number of jointed wavelengths enormously, simplify the model and increase the accuracy remarkably. SPA requires a less computational work load but GA is preferable in prediction accuracy. In conclusion, after spectral processing and the selection of training set, the GA-PCR model can accurately predict soil OMC while occupies fewer computational resources. The proposed method,i.e., the adoption of wavelength optimization before regression, can provide a new thought for the practical application of Vis-NIRS in the prediction of soil properties.

    [1] BEN-DOR E, BANIN A. Near infrared analysis as a rapid method to simultaneously evaluate several soil properties [J].SoilSci.Soc.Am.J., 1995, 59:364-372.

    [2] REEVES J B, MCCARTY G W, Meisinger J J. Near infrared reflectance spectroscopy for the determination of biological activity in agricultural soils [J].JNIRS, 2000, 8(3):161-170.

    [3] REEVES J B, MCCARTY G W, REEVES V B,etal.. Mid-versusnear-infrared diffuse reflectance spectroscopy for the quantitative analysis of organic matter in soils and other biological materials [J].AbstractofPapersofAmericanChemicalSociety, 2002, 223:U141-U142.

    [4] DUNN B W, BEECHER H G, BATTEN G D,etal.. The potential of near-infrared reflectance spectroscopy for soil analysis—a case study from the Riverine Plain of south-eastern Australia [J].Aust.J.Exp.Agr., 2002, 42(5):607-614.

    [5] SHEPHERD K D, WALSH M G. Development of reflectance spectral libraries for characterization of soil properties [J].SSSA, 2002, 66(3):988-998.

    [6] ISLAM K, SINGH B, MCBRATNEY A B. Simultaneous estimation of various soil properties by ultra-violet, visible and near-infrared reflectance spectroscopy [J].Aust.J.SoilRes., 2003, 41:1101-1114.

    [7] BEN-DOR E, IRONS J, EPEMA G F.SoilReflectance:RemoteSensingforTheEarthScience[M]. New York: John Wiley & Sons Inc., In: RENCZ A N. (Ed.), 3rd. Manual of Remote Sensing, 1999.

    [8] VASQUES G M, GRUNWALD S, SICKMAN J O. Comparison of multivariate methods for inferential modeling of soil carbon using visible/near-infrared spectra [J].Geoderma, 2008, 146:14-25.

    [9] VASQUES G M, GRUNWALD S, SICKMAN J O. Modeling of soil organic carbon fractions using visible?near-infrared spectroscopy [J].SoilSci.Soc.Am.J., 2009, 73:176-184.

    [10] MOUAZEN A M, KUANG B, DE BAERDEMAEKER J,etal.. Comparison among principal component, partial least squares and back propagation neural network analyses for accuracy of measurement of selected soil properties with visible and near infrared spectroscopy [J].Geoderma, 2010, 158:23-31.

    [11] STEVENS A, UDELHOVEN T, DENIS A,etal.. Measuring soil organic carbon in crop land satregional scale using airborne imaging spectroscopy [J].Geoderma, 2010, 158:32-45.

    [12] VISCARRA ROSSEL R A, BEHRENS T. Using data mining to model and interpret soil diffuse reflectance spectra [J].Geoderma, 2010, 158:46-54.

    [13] CHANG C W, DAVID A L, MAURICE J M,etal.. Near-infrared reflectance spectroscopy—principal components regression analyses of soil properties [J].SoilSci.Soc.Am.J., 2001, 65:480-490.

    [14] WANG Y B, HUANG T Y, LIU J,etal.. Soil pH value, organic matter and macronutrients contents prediction using optical diffuse reflectance spectroscopy [J].Comput.Electron.Agr., 2015, 111:69-77.

    [15] LIU L W. Formation and evolution of vertisols in the Huaibei Plain [J].Pedosphere, 1991, 1:3-15.

    [16] 李錄久,郭熙盛,王道中,等. 淮北平原砂姜黑土養(yǎng)分狀況及其空間變異 [J]. 安徽農(nóng)業(yè)科學(xué), 2006, 34 (4):722-723. LI L J, GUO X S, WANG D Z,etal.. State and spatial variability of nutrient of lime concretion black soil in Huaibei Plain [J].J.AnhuiAgric.Sci., 2006,34 (4):722-723. (in Chinese)

    [17] NATHAN M, GELDERMAN R.RecommendedChemicalSoilTestProceduresforTheNorthCentralRegion[M]. USA:North Central Regional Research Publication No. 221, 2012.

    [18] 魯如坤. 土壤農(nóng)業(yè)的化學(xué)分析方法 [M]. 北京:中國農(nóng)業(yè)科學(xué)出版社, 2000:106-107. LU R K.ChemicalAnalysisMethodofAgriculturalSoil[M]. Beijing: China Agricultural Science Press, 2000:106-107. (in Chinese)

    [19] RINAN A, VAN DEN BERG F W J, ENGELSEN S B. Review of the most common preprocessing techniques for near-infrared spectra [J].TrendsAnal.Chem., 2009, 28(10):1201-1222.

    [20] SAVITZKY A, GOLAY M J E. Smoothing and differentiation of data by simplified least squares procedures [J].Anal.Chem.,1964, 36(8):1627-1639.

    [21] WU W, WALCZAK B, MASSART D L,etal.. Artificial neural networks in classification of NIR spectra data: design of training set [J].Chemometr.Intell.Lab.Syst., 1996, 33(1):35-46.

    [23] 章海亮,劉雪梅,何勇. LS-SVM檢測土壤有機(jī)質(zhì)和速效鉀研究 [J]. 光譜學(xué)與光譜分析, 2014, 34(5):1348-1351. ZHANG H L, LIU X M, HE Y. Measurement of soil organic matter and available K based on SPA-LS-SVM [J].Spectrosc.Spect.Anal., 2014, 34(5):1348-1351. (in Chinese)

    [24] 陸婉珍. 現(xiàn)代近紅外光譜分析技術(shù) [M]. 北京:中國石化出版社, 2001:56-67. LU W Z.ModelNIRSpectroscopy[M]. Beijing, China Petro-chemical Press, 2001:56-67. (in Chinese)

    [25] ABDI H, WILLIAMS L J. Principal component analysis [J].WileyInterdisciplinaryReviews:Comput.Stat., 2010, 2(4):433-459.

    [26] MOUAZEN A M, BAERDEMAEKER J D, RAMON H. Effect of wavelength range on the measurement accuracy of some selected soil constituents using visual-near infrared spectroscopy [J].JNIRS, 2006, 14(1):189-199.

    林志丹(1981-),男,山東棲霞人,博士研究生,講師,2009年于合肥電子工程學(xué)院獲得碩士學(xué)位,主要從事光譜分析與建模等方面的研究。

    E-mail: linzd@mail.ustc.edu.cn王儒敬(1964-),男,安徽亳州人,博士,研究員,2004年于中科院合肥物質(zhì)科學(xué)研究院獲得博士學(xué)位,主要從事農(nóng)業(yè)智能系統(tǒng)的理論、方法與技術(shù)的研究。

    E-mail: rjwang@iim.ac.cn汪玉冰(1985-),女,安徽六安人,博士,副研究員,2010年于中科院合肥物質(zhì)科學(xué)研究院獲得博士學(xué)位,主要從事分子光譜分析及化學(xué)計量學(xué)在精準(zhǔn)農(nóng)業(yè)中的應(yīng)用等方面的研究。

    E-mail: ybwang@iim.ac.cn

    2015-05-30;

    2016-06-25

    中科院科技服務(wù)網(wǎng)絡(luò)計劃(KFJ-EW-STS-069)資助項目

    波長優(yōu)選對土壤有機(jī)質(zhì)含量可見光/近紅外光譜模型的優(yōu)化

    林志丹1,2,3, 汪玉冰1*, 王儒敬1*,汪六三1, 魯翠萍1, 張正勇1, 宋良圖1, 劉 洋1

    (1. 中國科學(xué)院 合肥智能機(jī)械研究所, 安徽 合肥 230031;2. 中國科技大學(xué) 自動化系, 安徽 合肥 230026; 3. 合肥電子工程學(xué)院, 安徽 合肥 230037)

    可見光/近紅外光譜模型是土壤屬性預(yù)測的有效工具。波長優(yōu)選在光譜建模過程中起著重要作用。文中首先利用從安徽省渦陽縣采集的130個砂姜黑土土壤樣本獲得可見光/近紅外光譜,然后利用平滑與多重散射校正聯(lián)合的光譜預(yù)處理方式消除光譜中的無關(guān)變量和冗余信息以提高模型預(yù)測結(jié)果的相關(guān)性,再利用SPXY方法挑選建模集樣本,分別利用連續(xù)投影算法和遺傳算法進(jìn)行波長優(yōu)選,最后利用留一法進(jìn)行交互驗證建立有機(jī)質(zhì)含量的主成分回歸模型。研究結(jié)果顯示:連續(xù)投影算法和遺傳算法都可以有效地減少參與建模的波長數(shù)并提高模型的準(zhǔn)確度,尤其是遺傳算法能夠更好地提高土壤有機(jī)質(zhì)含量預(yù)測精度,其相關(guān)系數(shù)、預(yù)測均方根誤差和相對分析誤差分別達(dá)到0.931 6,0.214 2和2.319 5。通過合適的特征波長選取,不僅計算量可以大大減少,預(yù)測精度也會有效提高。

    可見光/近紅外光譜; 有機(jī)質(zhì)含量; 光譜預(yù)處理; 樣本選擇; 波長優(yōu)化

    1000-7032(2016)11-1428-08

    O235 Document code: A

    10.3788/fgxb20163711.1428

    *CorrespondingAuthors,E-mail:ybwang@iim.ac.cn;rjwang@iim.ac.cn

    猜你喜歡
    中科院合肥波長
    HPLC-PDA雙波長法同時測定四季草片中沒食子酸和槲皮苷的含量
    合肥的春節(jié)
    加大授權(quán)力度中科院先行一步
    科技傳播(2019年23期)2020-01-18 07:57:10
    中科院沈陽生態(tài)研究所技術(shù)
    雙波長激光治療慢性牙周炎的療效觀察
    合肥:打造『中國IC之都』
    日本研發(fā)出可完全覆蓋可見光波長的LED光源
    中國照明(2016年4期)2016-05-17 06:16:15
    便攜式多用途光波波長測量儀
    物理實驗(2015年9期)2015-02-28 17:36:46
    中科院位列自然指數(shù)全球首位
    生態(tài)合肥
    天天添夜夜摸| 香蕉国产在线看| 亚洲成国产人片在线观看| 青青草视频在线视频观看| 真人做人爱边吃奶动态| 狠狠精品人妻久久久久久综合| 久久毛片免费看一区二区三区| 亚洲av日韩在线播放| 人人妻人人澡人人看| 50天的宝宝边吃奶边哭怎么回事| 侵犯人妻中文字幕一二三四区| 日本欧美视频一区| 人成视频在线观看免费观看| 午夜久久久在线观看| 热99国产精品久久久久久7| av在线app专区| 久久久国产精品麻豆| 久9热在线精品视频| 亚洲情色 制服丝袜| 亚洲欧美日韩另类电影网站| 精品人妻在线不人妻| 久久精品国产亚洲av高清一级| 极品人妻少妇av视频| 国产免费视频播放在线视频| 亚洲全国av大片| 黄色视频不卡| 精品国产乱码久久久久久小说| 91老司机精品| 午夜精品国产一区二区电影| 久久久水蜜桃国产精品网| 国产精品偷伦视频观看了| 免费av中文字幕在线| 又黄又粗又硬又大视频| 天天躁狠狠躁夜夜躁狠狠躁| 精品国产乱码久久久久久男人| 亚洲伊人久久精品综合| 两性午夜刺激爽爽歪歪视频在线观看 | 一级a爱视频在线免费观看| 国产日韩欧美在线精品| 亚洲国产欧美一区二区综合| 制服人妻中文乱码| 久久久久久人人人人人| 色94色欧美一区二区| 亚洲人成电影免费在线| 一区二区三区四区激情视频| 热99国产精品久久久久久7| 在线观看www视频免费| 久久国产精品影院| 一二三四在线观看免费中文在| 亚洲一区中文字幕在线| 亚洲五月色婷婷综合| 国产一级毛片在线| 午夜免费成人在线视频| 国产精品久久久av美女十八| 两个人免费观看高清视频| 精品乱码久久久久久99久播| 精品人妻熟女毛片av久久网站| 人人妻人人澡人人爽人人夜夜| 男女边摸边吃奶| 欧美黄色淫秽网站| 丰满少妇做爰视频| 日本91视频免费播放| 欧美激情高清一区二区三区| 国产亚洲午夜精品一区二区久久| 国产日韩欧美视频二区| www.av在线官网国产| 亚洲人成电影免费在线| 我要看黄色一级片免费的| 少妇精品久久久久久久| 中亚洲国语对白在线视频| 欧美国产精品一级二级三级| 亚洲精品国产区一区二| 色94色欧美一区二区| 亚洲国产av影院在线观看| 99久久99久久久精品蜜桃| 97在线人人人人妻| 黄频高清免费视频| www.熟女人妻精品国产| 国产成人欧美| 亚洲精品国产精品久久久不卡| 18禁黄网站禁片午夜丰满| 精品少妇内射三级| 精品人妻一区二区三区麻豆| 欧美在线一区亚洲| 国产成人av激情在线播放| 欧美亚洲日本最大视频资源| 最近中文字幕2019免费版| 国产区一区二久久| 不卡av一区二区三区| 男女无遮挡免费网站观看| 欧美精品一区二区大全| 久久综合国产亚洲精品| 97在线人人人人妻| 中亚洲国语对白在线视频| 91大片在线观看| 午夜两性在线视频| 人人妻人人澡人人爽人人夜夜| 亚洲免费av在线视频| 国产亚洲精品一区二区www | 国产无遮挡羞羞视频在线观看| 日本精品一区二区三区蜜桃| 操美女的视频在线观看| 一级a爱视频在线免费观看| 宅男免费午夜| 久久久精品国产亚洲av高清涩受| 欧美日韩精品网址| 国产一区二区三区综合在线观看| 国产精品亚洲av一区麻豆| 不卡av一区二区三区| 午夜福利乱码中文字幕| 免费高清在线观看视频在线观看| 美女福利国产在线| av欧美777| 亚洲人成电影观看| 咕卡用的链子| avwww免费| 欧美日韩成人在线一区二区| 亚洲国产欧美网| av在线老鸭窝| 视频在线观看一区二区三区| 看免费av毛片| 午夜影院在线不卡| 成人黄色视频免费在线看| 99国产精品99久久久久| 久热这里只有精品99| 91国产中文字幕| 啦啦啦在线免费观看视频4| 日本五十路高清| 狠狠狠狠99中文字幕| 欧美另类一区| 91字幕亚洲| 性色av乱码一区二区三区2| 亚洲精品国产一区二区精华液| 秋霞在线观看毛片| 成在线人永久免费视频| 成人三级做爰电影| 国产高清视频在线播放一区 | 五月天丁香电影| 他把我摸到了高潮在线观看 | 啦啦啦中文免费视频观看日本| av天堂在线播放| 在线看a的网站| 正在播放国产对白刺激| 91字幕亚洲| 欧美xxⅹ黑人| 人人妻人人添人人爽欧美一区卜| 精品亚洲成a人片在线观看| 大型av网站在线播放| 91大片在线观看| 男人舔女人的私密视频| 性色av乱码一区二区三区2| 伦理电影免费视频| 伊人久久大香线蕉亚洲五| 肉色欧美久久久久久久蜜桃| 搡老乐熟女国产| 天天躁夜夜躁狠狠躁躁| 日韩人妻精品一区2区三区| 黄色怎么调成土黄色| av线在线观看网站| 免费看十八禁软件| 亚洲欧美一区二区三区久久| h视频一区二区三区| 中文欧美无线码| 亚洲中文日韩欧美视频| kizo精华| 女人精品久久久久毛片| 精品国产超薄肉色丝袜足j| 午夜福利免费观看在线| 国产高清videossex| 在线亚洲精品国产二区图片欧美| 日韩人妻精品一区2区三区| xxxhd国产人妻xxx| 丝袜喷水一区| 亚洲一卡2卡3卡4卡5卡精品中文| 亚洲av片天天在线观看| 亚洲av美国av| 老司机午夜福利在线观看视频 | 国产深夜福利视频在线观看| 亚洲精华国产精华精| 精品国内亚洲2022精品成人 | 人妻 亚洲 视频| 1024香蕉在线观看| 国产在线免费精品| 汤姆久久久久久久影院中文字幕| 国产男人的电影天堂91| av电影中文网址| av线在线观看网站| 国产在线免费精品| 欧美日韩视频精品一区| 精品第一国产精品| 狂野欧美激情性bbbbbb| 日韩欧美一区视频在线观看| 国产男女超爽视频在线观看| 大香蕉久久网| 高清黄色对白视频在线免费看| 精品一区在线观看国产| 丝袜喷水一区| 女警被强在线播放| 下体分泌物呈黄色| 亚洲人成77777在线视频| 国产成人啪精品午夜网站| 少妇粗大呻吟视频| 亚洲av国产av综合av卡| 午夜福利影视在线免费观看| 日韩,欧美,国产一区二区三区| 狂野欧美激情性xxxx| 国产精品一二三区在线看| 黄色视频,在线免费观看| 2018国产大陆天天弄谢| 亚洲少妇的诱惑av| 精品熟女少妇八av免费久了| 欧美精品啪啪一区二区三区 | 成年人黄色毛片网站| 777久久人妻少妇嫩草av网站| 女人高潮潮喷娇喘18禁视频| 亚洲国产精品一区二区三区在线| 777米奇影视久久| 精品少妇一区二区三区视频日本电影| 女人久久www免费人成看片| 欧美日韩一级在线毛片| 在线观看免费高清a一片| 国产深夜福利视频在线观看| 免费观看av网站的网址| www.熟女人妻精品国产| 天天躁夜夜躁狠狠躁躁| 久久人妻熟女aⅴ| 两个人看的免费小视频| 欧美+亚洲+日韩+国产| 成年美女黄网站色视频大全免费| 可以免费在线观看a视频的电影网站| 久久久久网色| 精品人妻熟女毛片av久久网站| 99热网站在线观看| 97人妻天天添夜夜摸| 欧美黄色片欧美黄色片| 热99re8久久精品国产| 人妻人人澡人人爽人人| 美女午夜性视频免费| 麻豆av在线久日| 91老司机精品| 久久久久久久国产电影| 高清黄色对白视频在线免费看| 男女高潮啪啪啪动态图| 一区二区三区激情视频| 日韩熟女老妇一区二区性免费视频| 欧美黄色淫秽网站| 丰满少妇做爰视频| av欧美777| 成人黄色视频免费在线看| netflix在线观看网站| 少妇精品久久久久久久| 国产精品av久久久久免费| 大码成人一级视频| 亚洲欧美一区二区三区久久| 成人av一区二区三区在线看 | 最黄视频免费看| 两个人看的免费小视频| 国产无遮挡羞羞视频在线观看| 久久久精品免费免费高清| 又大又爽又粗| 亚洲国产精品一区三区| 啦啦啦啦在线视频资源| 精品少妇黑人巨大在线播放| 国产av又大| 久久久久久久精品精品| 老熟女久久久| 黄色视频,在线免费观看| 婷婷丁香在线五月| 欧美日韩av久久| 捣出白浆h1v1| 黄片播放在线免费| 国产伦理片在线播放av一区| svipshipincom国产片| 在线观看人妻少妇| 黄色片一级片一级黄色片| 午夜两性在线视频| 这个男人来自地球电影免费观看| 国产精品久久久久久精品电影小说| 国产精品自产拍在线观看55亚洲 | 欧美日韩亚洲国产一区二区在线观看 | 国产熟女午夜一区二区三区| 精品国产超薄肉色丝袜足j| 交换朋友夫妻互换小说| 欧美 亚洲 国产 日韩一| 久久精品人人爽人人爽视色| 真人做人爱边吃奶动态| 日本黄色日本黄色录像| 黄色怎么调成土黄色| 久久毛片免费看一区二区三区| 可以免费在线观看a视频的电影网站| 亚洲 国产 在线| 青草久久国产| 又紧又爽又黄一区二区| 午夜福利在线观看吧| 在线十欧美十亚洲十日本专区| 精品久久久久久久毛片微露脸 | 他把我摸到了高潮在线观看 | 黄色视频在线播放观看不卡| 久久精品国产亚洲av高清一级| 亚洲国产欧美网| 精品人妻熟女毛片av久久网站| av一本久久久久| 免费少妇av软件| 精品久久久久久电影网| 亚洲av成人不卡在线观看播放网 | 亚洲成人免费av在线播放| 久9热在线精品视频| 国产亚洲av高清不卡| 国产精品久久久久久人妻精品电影 | 两性午夜刺激爽爽歪歪视频在线观看 | 久久久久精品国产欧美久久久 | 精品国产乱码久久久久久小说| 一区二区三区激情视频| 午夜福利在线观看吧| 在线 av 中文字幕| 少妇裸体淫交视频免费看高清 | 欧美黑人欧美精品刺激| 欧美久久黑人一区二区| 久久av网站| 国产精品 欧美亚洲| 国产日韩欧美亚洲二区| 免费不卡黄色视频| 久久人人97超碰香蕉20202| 后天国语完整版免费观看| 欧美成人午夜精品| 大香蕉久久成人网| 动漫黄色视频在线观看| 后天国语完整版免费观看| 国产精品国产三级国产专区5o| 伊人亚洲综合成人网| 不卡av一区二区三区| 国产色视频综合| 免费在线观看视频国产中文字幕亚洲 | 欧美日韩亚洲国产一区二区在线观看 | svipshipincom国产片| 青春草视频在线免费观看| 最新在线观看一区二区三区| 天天躁狠狠躁夜夜躁狠狠躁| 精品国产一区二区久久| av天堂在线播放| 国产成人av教育| 女人被躁到高潮嗷嗷叫费观| bbb黄色大片| 女人被躁到高潮嗷嗷叫费观| 日本一区二区免费在线视频| www.自偷自拍.com| 亚洲欧美成人综合另类久久久| 久久久久久人人人人人| 三级毛片av免费| 久久久久久久国产电影| 一区二区av电影网| videos熟女内射| 日韩制服骚丝袜av| videos熟女内射| 免费不卡黄色视频| 国产成人系列免费观看| e午夜精品久久久久久久| 精品人妻一区二区三区麻豆| 亚洲美女黄色视频免费看| 黄网站色视频无遮挡免费观看| 亚洲精品久久久久久婷婷小说| 黑人巨大精品欧美一区二区mp4| 亚洲国产欧美一区二区综合| 秋霞在线观看毛片| 欧美精品一区二区免费开放| 每晚都被弄得嗷嗷叫到高潮| 女警被强在线播放| 女人高潮潮喷娇喘18禁视频| 爱豆传媒免费全集在线观看| 欧美97在线视频| 久久久国产一区二区| 精品一区二区三卡| www.熟女人妻精品国产| 亚洲av美国av| 一级毛片精品| 午夜福利在线免费观看网站| 日韩免费高清中文字幕av| 免费观看人在逋| 超碰成人久久| 国产精品 国内视频| 国产精品一区二区精品视频观看| 日韩人妻精品一区2区三区| 久热爱精品视频在线9| 亚洲精品日韩在线中文字幕| 另类精品久久| av福利片在线| av在线app专区| 两人在一起打扑克的视频| 国产日韩欧美亚洲二区| av网站在线播放免费| 男女无遮挡免费网站观看| 亚洲国产中文字幕在线视频| 午夜免费观看性视频| 欧美少妇被猛烈插入视频| 1024香蕉在线观看| av国产精品久久久久影院| 欧美激情极品国产一区二区三区| 99九九在线精品视频| 亚洲性夜色夜夜综合| 欧美日韩国产mv在线观看视频| 国内毛片毛片毛片毛片毛片| 亚洲午夜精品一区,二区,三区| 老鸭窝网址在线观看| 亚洲专区国产一区二区| 国产av又大| 国产欧美日韩一区二区三区在线| 两人在一起打扑克的视频| 国产精品麻豆人妻色哟哟久久| 精品欧美一区二区三区在线| 久久精品熟女亚洲av麻豆精品| 欧美国产精品一级二级三级| 欧美黑人欧美精品刺激| 美女福利国产在线| 狠狠婷婷综合久久久久久88av| 超碰成人久久| 中国美女看黄片| 99热全是精品| 久久香蕉激情| 精品人妻1区二区| 老司机午夜福利在线观看视频 | 国产在线视频一区二区| 大片免费播放器 马上看| 叶爱在线成人免费视频播放| 丝袜脚勾引网站| 最新的欧美精品一区二区| 亚洲中文字幕日韩| 一进一出抽搐动态| 搡老岳熟女国产| 欧美国产精品一级二级三级| 十八禁网站免费在线| 美女福利国产在线| 久久精品国产亚洲av高清一级| 欧美精品亚洲一区二区| 亚洲精品自拍成人| 蜜桃国产av成人99| 色婷婷av一区二区三区视频| 久久狼人影院| 黄片播放在线免费| 国产一区二区三区在线臀色熟女 | 最新的欧美精品一区二区| 国产精品久久久久久精品古装| 人人妻人人澡人人看| 男女午夜视频在线观看| 夫妻午夜视频| 日本精品一区二区三区蜜桃| 99久久99久久久精品蜜桃| 欧美xxⅹ黑人| 又紧又爽又黄一区二区| 亚洲人成电影免费在线| 欧美av亚洲av综合av国产av| 亚洲精品国产色婷婷电影| 亚洲久久久国产精品| 99国产精品99久久久久| 在线看a的网站| 欧美成人午夜精品| 五月开心婷婷网| 国产成人精品久久二区二区免费| 亚洲欧美精品自产自拍| 黑人巨大精品欧美一区二区mp4| 国产欧美日韩一区二区精品| 女人高潮潮喷娇喘18禁视频| 午夜久久久在线观看| 久久精品熟女亚洲av麻豆精品| 亚洲综合色网址| 午夜福利视频在线观看免费| 女人久久www免费人成看片| 欧美人与性动交α欧美精品济南到| 欧美精品啪啪一区二区三区 | 少妇人妻久久综合中文| 12—13女人毛片做爰片一| 国产av又大| 国产欧美日韩一区二区三区在线| 久9热在线精品视频| 久久久久久人人人人人| 国产欧美日韩综合在线一区二区| 飞空精品影院首页| 国产精品av久久久久免费| 青青草视频在线视频观看| 在线精品无人区一区二区三| 日日夜夜操网爽| 午夜福利免费观看在线| 亚洲色图综合在线观看| 90打野战视频偷拍视频| 国产免费一区二区三区四区乱码| 90打野战视频偷拍视频| 9191精品国产免费久久| 俄罗斯特黄特色一大片| 热re99久久国产66热| 五月开心婷婷网| 国产成人a∨麻豆精品| 亚洲中文av在线| 窝窝影院91人妻| 99国产精品免费福利视频| 一级毛片精品| 他把我摸到了高潮在线观看 | 中国国产av一级| 高清视频免费观看一区二区| 黄色a级毛片大全视频| 黄片小视频在线播放| 国产人伦9x9x在线观看| 91麻豆精品激情在线观看国产 | 热re99久久精品国产66热6| 成年人黄色毛片网站| 久久久久久免费高清国产稀缺| 两性夫妻黄色片| 黄色 视频免费看| 欧美日韩视频精品一区| 菩萨蛮人人尽说江南好唐韦庄| 天堂俺去俺来也www色官网| 免费高清在线观看视频在线观看| 久久久精品区二区三区| 亚洲精华国产精华精| 午夜激情av网站| 最近最新免费中文字幕在线| 精品福利观看| cao死你这个sao货| 丝袜脚勾引网站| 中亚洲国语对白在线视频| 亚洲五月色婷婷综合| 中国美女看黄片| 建设人人有责人人尽责人人享有的| 超色免费av| 另类精品久久| 亚洲一卡2卡3卡4卡5卡精品中文| 女人被躁到高潮嗷嗷叫费观| 免费在线观看黄色视频的| 91字幕亚洲| 菩萨蛮人人尽说江南好唐韦庄| 精品亚洲成a人片在线观看| 亚洲成国产人片在线观看| 啦啦啦免费观看视频1| 色综合欧美亚洲国产小说| 精品一品国产午夜福利视频| 夫妻午夜视频| 宅男免费午夜| 一级毛片女人18水好多| 美女高潮到喷水免费观看| 日日摸夜夜添夜夜添小说| 成人免费观看视频高清| 美国免费a级毛片| 永久免费av网站大全| 国产精品免费视频内射| 嫁个100分男人电影在线观看| 日本五十路高清| 久久影院123| 中文字幕另类日韩欧美亚洲嫩草| 性色av一级| 两人在一起打扑克的视频| 欧美激情 高清一区二区三区| 美国免费a级毛片| 中文欧美无线码| 国产日韩欧美在线精品| 一级毛片女人18水好多| cao死你这个sao货| 成人国产一区最新在线观看| 男女高潮啪啪啪动态图| 俄罗斯特黄特色一大片| 国产无遮挡羞羞视频在线观看| 亚洲成国产人片在线观看| www.999成人在线观看| 色婷婷久久久亚洲欧美| 久久精品久久久久久噜噜老黄| 亚洲精品久久成人aⅴ小说| 热re99久久国产66热| 久久久久国内视频| 亚洲成人免费av在线播放| 天堂8中文在线网| 国产免费福利视频在线观看| 成人av一区二区三区在线看 | 欧美黄色片欧美黄色片| 无限看片的www在线观看| 国产一区二区三区av在线| 巨乳人妻的诱惑在线观看| 日日摸夜夜添夜夜添小说| 1024视频免费在线观看| 在线观看人妻少妇| 精品少妇久久久久久888优播| 天天操日日干夜夜撸| 日韩欧美国产一区二区入口| 丰满迷人的少妇在线观看| 精品免费久久久久久久清纯 | 建设人人有责人人尽责人人享有的| 国产成人欧美在线观看 | 亚洲视频免费观看视频| 无限看片的www在线观看| 亚洲一卡2卡3卡4卡5卡精品中文| 久热这里只有精品99| 91九色精品人成在线观看| 亚洲色图 男人天堂 中文字幕| 美女中出高潮动态图| 91精品国产国语对白视频| 日韩人妻精品一区2区三区| av在线app专区| 精品少妇久久久久久888优播| 母亲3免费完整高清在线观看| 99国产极品粉嫩在线观看| 91老司机精品| 伊人亚洲综合成人网| 涩涩av久久男人的天堂| 黄色怎么调成土黄色| a级毛片在线看网站| 精品亚洲成a人片在线观看| 一级毛片女人18水好多| 久久天躁狠狠躁夜夜2o2o| 免费黄频网站在线观看国产| 窝窝影院91人妻| a在线观看视频网站| 国产真人三级小视频在线观看| 黄色毛片三级朝国网站| 亚洲精品久久久久久婷婷小说| 啦啦啦免费观看视频1| 大片免费播放器 马上看| tocl精华| 欧美变态另类bdsm刘玥| 欧美性长视频在线观看|